Text-to-Speech

What is it?

Text-to-speech on Liforma is the same Avatar Experience runtime with audio output only. You use the normal integration surface — <Experience> or Experience.startSession() — and set speechOnly so the player runs TTS (and optional STT) without loading avatar assets, location scenes, or the visual renderer.

For scripted lines, use presenter mode and call speak() after the session has started (typically from onStarted).

When to use it

• Voice prompts, IVR-style messages, or accessibility narration
• Apps that already have their own UI and only need spoken output
• Scripted lines via speak() in presenter mode
• When lip-sync, facial animation, and scene backgrounds are not required

Svelte

Add speechOnly to <Experience>. Use presenter mode with speechInputMode="off" when the host drives every line with speak().

<script lang="ts">
  import {
    Experience,
    type ExperienceHandle
  } from '@liforma/client/svelte';

  let experience: ExperienceHandle | undefined = $state();

  async function speakPrompt() {
    await experience?.speech.speak({ text: 'Your table is ready.' });
  }
<\/script>

<Experience
  bind:this={experience}
  experienceId="exp_T0I7ACMQLBMPG6K"
  speechOnly
  mode="presenter"
  speechInputMode="off"
  onStarted={speakPrompt}
/>

JavaScript

import { Experience } from '@liforma/client';

const experience = await Experience.startSession({
  experienceId: 'exp_T0I7ACMQLBMPG6K',
  speechOnly: true,
  mode: 'presenter',
  speechInputMode: 'off'
});

experience.on('started', async () => {
  await experience.speech.speak({ text: 'Your table is ready.' });
});

await experience.attach({ container: '#voice-shell' });

The hosted player still mounts in your container for audio playback and session UI (start button, close control). It simply skips avatar and location image loading.

Web component

Set speech-only="true" on <liforma-experience> with the same presenter + speak() flow. See Web component.

When to use a visible character instead

Omit speechOnly (or set it to false) and use the same experience with presenter mode and speak() when speech should come from an on-screen animated character.